As the density of VLSI circuits has increased, the circuit patterns themselves continue to be miniaturized. This miniaturization of circuit patterns has in turn engendered higher numerical apertures (NA) for the miniaturized projection lens systems used in semiconductor exposure apparatuses, such that the permissible depth of focus of the lens during the process of transferring the circuit pattern has become shallower. At the same time, the area to be exposed by the miniaturized projection lens system during the manufacture of a semiconductor chip has increased. Given these conflicting trends, there is a growing need for a scanner-type semiconductor exposure apparatus capable of enlarging the exposure area and NA while still using a stepper-type lens equivalent.
In a scanner-type exposure apparatus, a common method used to make a good exposure transfer of a circuit pattern across the entire exposure area involves scanning the wafer surface that is to receive the transfer of the circuit pattern on a reticle (hereinafter sometimes referred to as the focus plane of the wafer surface of the miniaturized projection lens system) so as to detect the position and tilt of the wafer surface and continuously adjusting the position and tilt of the focus plane via an auto-focus/auto-leveling correction drive operation so as to continuously present the best possible imaging surface to the projection lens.
There are several commonly known methods of detecting the height of the wafer and the position of the wafer surface (i.e., the position and tilt of the wafer surface described above) in such an arrangement. One commonly known method involves using a detection optical system to direct a beam of light onto the wafer surface at an angle thereto and to detect a positional deviation in the reflected light as a positional deviation on a sensor. Another commonly known method involves using a gap sensor, such as an air micro-sensor or an electrostatic capacitance sensor, to scan a plurality of points on the surface of the wafer and determine the exact position of the wafer surface from that scan.
Further, in the methods described above, in order to position the entire wafer exposure area (also called a “shot”) securely within the permissible depth of focus of the miniaturizing projection lens system—a depth of which, as described above, has become shallower with advances in the NA—the whole process is closely controlled so as to avoid the local topography beneath the point of scanning detection (or of light reflection), which can give erroneous readings, by using a plurality of positions within the shots as detection points, comparing the readings at the detection points against an optimum focus setting surface, measuring the difference (if any), and offsetting it.
Use of this sort of measurement offset is accomplished by one of two methods. The first method involves measuring the measurement offset itself using a pilot wafer to manage the job. The second method involves using the first wafer after a change in the wafer production lot to obtain the measurement offset, which is then used to process that lot.
However, the conventional apparatuses and methods have several problems, which are detailed below.
The scanning speed of a slit-scan-type exposure apparatus is calculated from the apparatus intensity (that is, the illumination intensity within the exposure slit defined by the masking plate) and the wafer surface resist sensitivity so as to yield a best exposure amount.
The intensity of the apparatus can be controlled by the laser output and the laser oscillation frequency, and is set to make the best exposure amount constant in the face of process discrepancies. Here, the term “process discrepancies” means the phenomenon in which the best exposure amount fluctuates within the same wafer due to unevenness in the resist coating on the wafer surface or unevenness in the high-speed dispersion furnaces used in the after-process baking of the wafers (wafer surface oxidation).
The scanning speed and the intensity of the apparatus are interrelated, insofar as, for a given wafer surface resist sensitivity, the scanning speed increases when the laser output and the laser oscillation frequency increase. Conversely, the scanning speed decreases when the laser output and the laser oscillation frequency decrease.
In other words, in the conventional art, a fixed scanning speed can be made constant for the same job or in the same lot by adjusting the best exposure amount using the laser output and the laser oscillation frequency to account for process discrepancies, with the above-described measurement offset obtained for the pilot wafer or the lead wafer in the production used in the processing of that lot.
However, since the scanning speed calculated in such a case becomes the lowest of the speeds among which it is possible to select with respect to the process discrepancies, a problem arises in the throughput. The problem is magnified in the case of DRAMs, which tend to have large individual production lots.
Also, when using low-output laser light source such as Argon-Fluorine (ArF) lasers and F2 lasers, the intensity of the apparatus is insufficient relative to the scanning speed, which is reduced accordingly. As a result, when attempting to control for the best exposure amount using the laser output and the laser oscillation frequency for the process discrepancies, the lack of scanning speed makes improvements in throughput difficult to achieve.
Moreover, even if the scanning speed is increased in order to improve the throughput, the insufficiency of apparatus intensity relative to the scanning speed makes it difficult to control the best exposure amount using the laser output and the laser oscillation frequency in the face of process discrepancies.
One measure used to solve the above-described problem involves controlling the best exposure amount for the process discrepancies by changing the scanning speed with each shot and each wafer, thereby making maximum use of the apparatus intensity. However, the time allotted for the operation of detecting and measuring the position and tilt of the wafer surface while being synchronized with the scanning operation is fixed, so the scan area will vary from shot to shot and wafer to wafer if the scanning speed is changed with each shot and each wafer.
If, in such circumstances, the above-described conventional control method are used to carry out the measurement offset described above, the size of the scan area will be different for the measurement offset and the exposure, leading to an error in the measurement offset for the optimum focus setting surface.
Conversely, seeking the measurement offset for each scanning speed, although it does not lead to errors in the measurement offset for the optimum focus setting surface, does reduce throughput drastically due to the great frequency with which opportunities arise to seek the measurement offset.